Four success stories of skills-based volunteering at Ersilia
We have found that skill-based volunteering is an effective strategy for enabling open source contributions in biomedical research from software developers who, despite not specialising in this field, are keen to participate
The idealised view of an open-source project is one of organically growing software, self-managed and fueled by the input of a diverse group of global contributors. These contributors are driven by the mission, urgency, and impact of the project, their eagerness to learn from each other, and a shared belief in code as a common good. Although there are more refined definitions of open source, this perspective influenced the expectations for the Ersilia Model Hub, our flagship project, when we launched the initiative three years ago.
However, our experience with this decentralised, somewhat ungoverned approach has been less successful than we had hoped. Several factors may have contributed to this, the most significant being the steep learning curve required to contribute meaningfully to Ersilia, given our narrow field of action (AI for antimicrobial drug discovery) and, simply put, the fact that our developers (software engineers) are not our users (scientists). This complexity makes onboarding developers difficult, especially when contributions are intermittent and voluntary. It now seems overly optimistic to expect a project like Ersilia to grow autonomously and naturally in response to the ‘needs’ of drug discovery research without a clear, intentional plan and a set of discrete tasks outlined by a scientific leadership.
Instead, over time, we have identified two effective mechanisms for fostering significant open source contributions. The first is a series of 12-week internship rounds, notably including the Outreachy program, where 3–4 interns become integral members of our daily team operations. We strive to mentor them and, importantly, they are sponsored and find in this experience a gateway to the broader market. Generally, interns are tasked with increasing Ersilia’s AI model count, each handling models of escalating complexity and leading to substantial code contributions in the public domain, which is good for their profile. With nearly 150 models in our platform, this method has proven effective. Often, interns belong to underrepresented minorities in STEM, so training them is very much aligned with our mission.
The second mechanism involves teams from organisations, typically from the global north, sponsored or supported by their employers to volunteer with us. This arrangement is unique: we define a project and, for a set period, the volunteers approach the task as they would approach a contracted service. This collaboration allows us to expand our code infrastructure, particularly through the development of professional-looking workflows, databases and protocols. Unlike intern cohorts, these volunteer teams have a lead or project manager themselves, responsible for coordination, project planning, execution and communication with us. Ersilia’s primary role is not to mentor these volunteers but to ensure they have an impactful project to work on. They work autonomously, seeking feedback as needed, presenting progress, and ultimately delivering a well-documented, mature product. These engagements are incredibly educational for us, and we often ponder what we can offer in return beyond our gratitude and commitment to effectively utilise and honour their contributions. As 2023 draws to a close, I wish to publicly acknowledge the exceptional work of four skill-based volunteer teams over the year.
GitHub Social Impact
Ersilia, like many other open-source projects, relies heavily on GitHub. All our code resides on GitHub, so it seemed logical to start by seeking support from this company. Fortunately, the Social Impact team at GitHub assembled a group of volunteers who have truly transformed how we use their platform. They developed a set of GitHub Actions workflows that enable AI model contributions from our interns, which has not only helped us scale up but also relieved us of a significant burden in code revision and maintenance. A few months ago, we received the GitHub For Social Good Award in San Francisco during the GitHub Universe 2023 event. Additionally, as part of this collaboration, we presented in person at RightsCon Costa Rica, a highlight of the year for us.
The Good Data Institute
A different yet equally important example of skill-based volunteering at Ersilia is the Good Data Institute (GDI). GDI offers data science and engineering support to non-profit organisations and charities. We approached GDI with a tangible and technically challenging project related to mass calculations across the Ersilia Model Hub. We had been struggling with this issue for over a year without success. GDI delivered a carefully engineered and well-documented workflow, which will now enable us to advance our research agenda. From GDI, we learned the importance of thoughtful and iterative project design. We treasure their diagrams and technical documents as some of the best in our archives.
Digital Ocean Hollie’s Hub for Good
As an underfunded organisation, we have always hesitated to fully embrace cloud infrastructure due to our inability to manage the fluctuating and often unpredictable costs of operations. In this context, the clarity and in-kind support provided by Digital Ocean have been instrumental in making our journey into the cloud a bit easier, offering deployment solutions that have greatly facilitated remote scientific collaborations for us. Thanks to the volunteers from Digital Ocean, we are now able to deploy AI models online with just one click, a feat that seemed unattainable just a year ago.
Harvard Tech For Social Good
Finally, another highly successful collaboration we’ve had is with Harvard Tech for Social Good (T4SG). While managed by and for students, these are not internships per se; instead, T4SG volunteers execute a project end-to-end, led by a senior student who coordinates the rest. I am personally amazed by the two rounds of T4SG volunteers so far. In the first round, they developed a front-end for our Ersilia Model Hub (which I hope we can soon bring to production), and in the second round, they developed a dashboard for monitoring Ersilia using Splunk. We’ve learned a lot from the T4SG teams; I admire their autonomy and eloquence in delivering results.
A common factor underlying these four projects with highly qualified volunteer teams is that, as Ersilia, we need to ensure they are incentivized and always understand the motives behind their contributions. These teams possess skills beyond ours, and we cannot guide them on a day-to-day basis. Instead, what works really well is preparing and investing the necessary time and resources in project planning, offering them a clear view of the bigger picture and the scope of our research roadmap. We hope to continue with more volunteer engagements like these in 2024. So far, this model of open source contribution has proven to be very effective for a small open science organisation like ours.
